Ponente
Descripción
Our work is an attempt to understand the competitive aspects of the diffusion of ideas, innovations, or product adoption in social networks. Grounded in non-cooperative game theory, we propose a framework called "I-Game" (Influence Game), that models the adoption of competing products in social networks as a strategic game in which the firms are the players. The finite strategy space is the set of algorithms for selecting their campaign initiators (seeds) and their payoff is the expected number of adopters of their respective products when diffusion occurs according to the Competitive Independent Cascade (IC) model. The competitors simultaneously decide which potential seeds to choose to obtain the largest cascade of their product adoption. We observe through intensive simulations that when considering a finite strategy space provided by the classical centrality measures and by the most effective propagation-based diffusion algorithms without competition, the former strategies outperform the latter ones in terms of spread in a competitive diffusion setting. To explain this behaviour, we introduce a new measure, Initial Diffusion Velocity, that gauges the diffusion spread over time, which is a useful metric to predict the outcome of competitive diffusions in SN. In this work, on the basis of the obtained results, we propose the use of hybrid strategies, that selects the potential seed sets provided by two algorithms, the former seeds are selected by an algorithm that is fast, while the latter are selected according to slower, but more effective algorithms in non-competitive models.
